The standard benefit incidence algebra generally produces biased estimates of the distribution of public spending on education when students from poor and rich families are enrolled in schools that receive different levels of public spending per student. Except in very rare instances, removing these biases entails combining several sources of information in order to evaluate how unit spending varies across different population groups. Although such disaggregation is generally difficult to obtain, we show one way to overcome the data constraints that hinder a precise calculation of the incidence of public spending on education. The empirical example discussed in this article indicates that after disparities in the level of public spending per student are controlled for, the results can be quite different. (Contains 3 tables and 4 footnotes.)